Indo-European Reduplication: Synchrony, Diachrony, and Theory by Sam Zukoff M.A., University of Georgia (2012) B.A., Georgetown University (2010) Submitted to the Department of Linguistics and Philosophy in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Linguistics at the MASSACHUSETTS INSTITUTE OF TECHNOLOGY September 2017 c Sam Zukoff, MMXVII. All rights reserved. The author hereby grants to MIT permission to reproduce and to distribute publicly paper and electronic copies of this thesis document in whole or in part in any medium now known or hereafter created. Author........................................................................... Department of Linguistics and Philosophy September 8th, 2017 Certified by....................................................................... Donca Steriade Professor Thesis Supervisor Accepted by...................................................................... David Pesetsky Head, Department of Linguistics and Philosophy
Indo-European Reduplication: Synchrony, Diachrony, and Theory by Sam Zukoff Submitted to the Department of Linguistics and Philosophy on September 8th, 2017, in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Linguistics Abstract The reduplicative systems of the ancient Indo-European languages are characterized by an unusual alternation in the shape of the reduplicant. The related languages Ancient Greek, Gothic, and Sanskrit share the property that root-initial consonant clusters exhibit different reduplicant shapes, depending on their featural composition. Moreover, even though the core featural distinction largely overlaps across the languages, the actual patterns which instantiate that distinction are themselves distinct across the languages. For roots beginning in stop-sonorant clusters (TRVX roots), each of these languages agrees in displaying a prefixal CV reduplicant, where the consonant corresponds to the root-initial stop: TV-TRVX. These three languages likewise agree that roots beginning in sibilant-stop clusters (STVX roots) show some pattern other than the one exhibited by TRVX roots. However, each of the three languages exhibits a distinct alternative pattern: V-STVX in the case of Ancient Greek, STV-STVX in the case of Gothic, TV-STVX in the case of Sanskrit. This dissertation provides an integrated synchronic and diachronic theoretical account of the morphophonological properties of verbal reduplication in the ancient Indo-European languages, with its central focus being to explain this core alternation between TRVX roots and STVX roots. Set within Base-Reduplicant Correspondence Theory, a framework for analyzing reduplication in Optimality Theory, the comprehensive synchronic analyses constructed in service of understanding this distinction and other interrelated distinctions allow us to probe complex theoretical questions regarding the constraints and constraint interactions involved in the determination of reduplicant shape. This dissertation seeks not only to develop in depth, consistent accounts of both the productive and marginal/archaic morphophonological aspects of reduplication in the Indo-European languages, it aims to understand the origins of these patterns from a historical and comparative perspective, and from the perspective of morphophonological learning and grammar change and attempts to motivate the conditions for the onset, development, and retention of the changes that result in the systems observed in the attested languages. As such, these analyses constitute a valuable set of case studies on complex systemic change in phonological grammars. Thesis Supervisor: Donca Steriade Title: Professor 3
To perseverance, and to everyone who helped me along the way. 5
Acknowledgments They say that the dissertation is not the end, it s just the beginning. Well, whatever this is, it began well before I ever conceived of writing a dissertation. This whole giant mess emerged out of a phonology assignment sophomore year of college at Georgetown. In Jim Gruber s phonology class, we had to come up with our own problem set. I happened to be taking my first semester of Ancient Greek at the time, and I had noticed that this whole reduplication thing was kind of weird. So I used that for my problem set, which I eventually decided to turn into my senior thesis (and no, I m not going to let anyone see it), which morphed into my class paper for Anya Lunden s graduate phonology class at UGA, which morphed into my UGA Master s thesis (which I might let you see if you ask nicely), which prompted my independent study with Donca Steriade at MIT, which I turned into my first conference presentation at WECIEC, for which I wrote my first proceedings paper, which turned into my first journal article at LI, which forms the basis for the first content chapter of this dissertation, which is entirely an extension of the ideas contained therein. So many people have helped me get from there to here, and even to get there in the first place. I can t possibly thank them all here, let alone thank them properly. But here s my best attempt. First I have to thank my committee, who have all been extremely supportive throughout the dissertation process, and through my whole time at MIT. To Donca Steriade, a great advisor and my greatest champion. When I first arrived on campus for my accepted student visit, I stepped off the elevator and saw the department and said to myself Wow, this is different, and awesome!. Then the very first thing I did was meet with Donca, and I had the very same reaction. I made the department wine me and dine me for the rest of the open house, but I knew at that moment that I was going to MIT. Thank you, Donca, for always being willing to read and comment on everything I ever wrote, no matter how bad a shape it was in. And for always being open to my ideas, even (and especially) when you adamantly disagreed with them. For giving me the chance to argue for my ideas, and for be willing to be convinced when I did a good job at doing so. I can think of no greater compliment than, when after my second generals defense, you came to me and said This is the second time you came to me with something crazy... and it turned out to be true!. There are so many different possibilities that I can only guess about which was the first time she had in mind. To Adam Albright, for always having the most positive and constructive feedback. I can t count the number of times I went into a meeting with you feeling really anxious and uncertain about how to proceed on a project and whether I would ever be able to make any headway on it, and came out with a game plan that I was confident about and believing that I could do it. And for, like clockwork, always asking the question But how would you learn this?. And to Edward Flemming, for never being satisfied with anything less than the absolute clearest and strongest reasoning. I came out of every meeting with you with a clearer understanding of my own thoughts, and of the bigger picture. And thank you to all the rest of the faculty and staff at MIT Linguistics, as well, especially David Pesetsky and Michael Kenstowicz. There were multiple times when you all went above and beyond to help me through difficult times, and your kindness and support made a huge difference. I ve also been extremely lucky to have support from giants of the field who weren t on my committee. There was no greater validation than when Craig Melchert read my Anatolian chapter and gave it the thumbs up. Thank you, Craig, for suggesting that I look into Anatolian in the first place, for being willing to read and comment on my chapter (and for doing so so quickly), and for, as always, providing the most helpful feedback. Though I didn t take full advantage of it, having Jay Jasanoff down the street has been an amazing resource for me. Thank you, Jay, for be willing to 6
meet with me any time I asked, and for me indulging me in my ideas even when they were counter to your own. Your feedback on the Germanic chapter was unbelievably helpful. Thank you to both of you for allowing me to be an honorary member of UCLA Indo-European Studies and Harvard Linguistics, respectively. As I said before, this work got started long before I ever thought I had a chance of coming to MIT (or would even want to). There s no chance I would have made it here if not for my amazing Master s advisors at UGA: Jared Klein and Anya Lunden. Everything I know about Indo-European I know because of Jared Klein. Jared took me in at UGA when it looked like I had no other options. Without his generous support and his guidance, I never would have reached this point. And thank you to Anya, for giving me my first real taste of theoretical phonology, and getting me hooked. Most of my aesthetic preferences in phonology can be traced back to what you taught me. I never would have been in a position to succeed at UGA without the great education I received at Georgetown. Thank you to Shaligram Shukla, Alex Sens, Lisa Zsiga, Patricia Slaton, Jim Gruber, and many others. And I never would have been in position to succeed at Georgetown without the amazing support and guidance I received at Summit High School. Thank you to Mr. Thayer, Mrs. Solondz, and Dr. Schnedeker I really wouldn t have made it through that time without you. I have been extremely lucky to work on various aspects of this project with some amazing coauthors, who happen to be some of my best friends. The fact that there s a chapter in this dissertation about Anatolian is completely due to Tony Yates. Without Tony, I never would have been able to make heads or tails out of that data, and so many of the detailed arguments from the historical phonology of those languages are due totally to him (except whichever ones might ultimately be judged not to work, which were all my ideas). I ve known Tony since Jared Klein designated him as the person to show me around on my visit to UGA. Getting to make our triumphant return to Athens to present this together at ECIEC couldn t have been more perfect. And thank you for making UCLA, and LA in general, like a second home for me these last 6 years. You and Sam always made me feel so at home and were always happy to let me overstay my welcome. I wouldn t trade all those brewery visits, trips to the beach in Santa Monica, and overall great company, for anything. The basis for the Germanic chapter came about when Ryan Sandell and I realized we were accidentally working on pretty much the same idea at the same time. We decided to join forces, and that couldn t have worked out any better. Ryan has been perhaps the most enthusiastic adopter of my ideas, and those ideas would never have reached the level where they are now without the benefit of his work and his insights. And thank you, Ryan, for your amazing feedback on drafts of several chapter, which were immeasurably improved from your suggestions. The work on *PCR immensely benefited from joint experimental work with Benjamin Storme. While the vicissitudes of dissertation writing prevented much of it from actually making it in, Ben s expert experimental acumen helped me conceptualize the problem in different and more precise ways. I look forward to continuing to work on this with you in the future. And thank you for yours and Liz s friendship over this time at MIT. Getting to spend a week having you show me around France, and getting to return the favor and show you around New Jersey and New York, are some of my fondest experiences in grad school. While we didn t work on this project together, my collaboration with Juliet Stanton has certainly made me a better phonologist. Juliet has constantly set the bar for me for how hard you should work and how rigorous your work should be. Thank you, Juliet, for always being there to talk about phonology (especially to whisper back and forth during talks) and always being interested in what I had to say. Thank you also to Tyler Lau. Tyler helped me run down a potentially relevant pattern in the Ryukyuan languages. I ultimately had to leave it out of the dissertation, but I really appreciate your help. 7
I don t think I ve ever had as much fun as I have with my friends from MIT. Thank you to Ted Levin and Coppe van Urk for always being willing to ball out of control, or, when circumstances dictated, to ball on a budget. Thank you to Ruth Brillman and Chris Davis for being the most supportive friends you could ever ask for, always inviting me over for a gourmet meal at the drop of a hat, and dropping everything to go grab a beer whenever I needed to talk. Michelle Yuan, Athulya Aravind, and Carrie Spadine insisted that they deserved special mention too. They re right. Thanks also to (in chronological order) Sam Steddy, Gretchen Kern, Aron Hirsch, Anthony Brohan, Chris O Brien, Kenyon Branan, Ezer Rasin, Justin Colley, Colin Davis, Chris Baron, and all the other amazing friends I ve had the honor of sharing the department with (and to Nicole Torres, who, at this point, is basically an honorary member of the department). Thanks to my linguist friends from near and far: Nico Baier, Laura Grestenberger, Jesse Lundquist, Chiara Bozzone, Andrew Bird, Jess DeLisi, Caley Smith, Hemanga Dutta, and too many others to count. And thanks also to my great friends from back in Summit, New Jersey: Dan Rufolo, David Richards, Scott Leighton, just to name a few. Thank you also to my Boston area cousins: Steve Zukoff, Adam and Beth Badik, Kenny Westerman and Brian Westerman. Above all, however, I could not have made it to this point without the unwavering and unparalleled love and support from my family. To my parents, Mimi and Paul Zukoff, words can t describe how grateful I am for everything you ve done for me, and everything you continue to do for me. I never would have made it through all the tough times without you in my corner. Thank you to my grandmother Irene Preiser and my late grandfather Aaron Preiser. You were always there to support me and facilitate my success. You know that I am who I am because I take after you. Thank you to my grandmother Ella Zukoff, my grandfather Marty Zukoff, and my uncle Ed Zukoff. No grandparents have ever shown their grandson as much love as you have always showed me. And thank you to my late brother Ben Zukoff. You have always been an inspiration to me, while you were and every day since. The incomparable Ted Levin once imparted us with the following wisdom: You can sweat the petty stuff, but don t pet the sweaty stuff. On this point I can quibble with only one thing: you shouldn t sweat the petty stuff either. To anyone who s still reading at this point, I must strongly recommend you stop now it s all downhill from here. But if you insist, here it is. 8
Contents 1 Introduction 15 1.1 Overview of the Dissertation............................. 15 1.2 Structure of the Dissertation............................. 16 1.2.1 Ancient Greek................................ 16 1.2.2 Anatolian................................... 17 1.2.3 Gothic..................................... 18 1.2.4 Sanskrit.................................... 18 1.2.5 The NO POORLY-CUED REPETITIONS Constraint (*PCR)......... 19 1.2.6 Reconstructing Proto-Indo-European Reduplication............. 20 1.3 Analysis of Reduplicant Shape in Indo-European.................. 21 1.3.1 Across-the-board Behavior.......................... 21 1.3.1.1 Across-the-board Cluster-copying: Hittite............ 21 1.3.1.2 Across-the-board C 1 -copying: Old Irish (and elsewhere)..... 24 1.3.1.3 Across-the-board C 2 -copying: Unattested............. 25 1.3.2 Cluster-Dependent Copying Patterns..................... 25 1.3.2.1 TRVX C 1 -copying, STVX Cluster-copying: Gothic...... 26 1.3.2.2 TRVX C 1 -copying, STVX C 2 -copying: Sanskrit........ 27 1.3.2.3 TRVX C 1 -copying, STVX Non-copying: Ancient Greek... 29 1.3.3 Factorial Typology.............................. 31 2 Greek 33 2.1 Introduction...................................... 33 2.1.1 A Preview of the Data............................ 34 2.1.2 Outline of the Chapter............................ 35 2.2 Reduplication in Ancient Greek........................... 36 2.2.1 Consonant-Initial Roots........................... 36 2.2.1.1 Data and Generalizations..................... 36 2.2.1.2 The Analysis of Reduplication in Base-Reduplicant Correspondence Theory........................... 37 2.2.1.3 Perfect Reduplication: One Morpheme or Two?......... 39 2.2.1.4 The C 1 -Copying Pattern...................... 40 2.2.1.5 The Non-Copying Pattern..................... 43 2.2.1.6 Local Summary.......................... 46 2.2.2 Vowel-Initial Roots.............................. 46 2.2.2.1 Vowel-Lengthening Perfects.................... 46 2.2.2.2 Attic Reduplication in Ancient Greek............... 48 9
2.2.3 REDUP(RED) lex, Reduplicated Presents, and their Associated Perfects... 50 2.2.3.1 The Reduplicated Presents.................... 50 2.2.3.2 Reduplicated Presents and Pre-Greek............... 51 2.2.3.3 Generating the Unexpected C 1 -Copying Presents and Perfects with REDUP(RED) lex....................... 52 2.2.4 Interim Summary............................... 53 2.3 Attic Reduplication.................................. 55 2.3.1 Attic Reduplication and the Laryngeals................... 55 2.3.2 Previous Approaches............................. 57 2.3.3 Vowel Prothesis and Laryngeal Vocalization in Greek............ 58 2.3.4 Generating (Pre-)Attic Reduplication in Pre-Greek............. 60 2.3.4.1 Motivating the Pattern....................... 60 2.3.4.2 The Alternative Pattern...................... 61 2.3.5 Attic Reduplication for *HeC roots..................... 64 2.3.6 Interim Conclusions............................. 65 2.4 The Diachrony of Attic Reduplication........................ 65 2.4.1 Compositionality in Greek Reduplication.................. 66 2.4.2 The Diachrony of Laryngeal Loss and the Reflexes of the Perfect..... 67 2.4.3 Inconsistency Detection and Constraint Cloning for Attic Reduplication.. 69 2.4.4 Deriving the Behavior of REDUP(RED) lex After the Rise of *PCR..... 73 2.4.5 Local Summary and Discussion....................... 75 2.5 Attic Reduplicated Presents and Aorists....................... 76 2.5.1 Attic Reduplicated Presents......................... 76 2.5.2 Attic Reduplicated Aorists.......................... 77 2.5.3 Consonant-initial Reduplicated Aorists................... 79 2.5.4 Local Summary................................ 81 2.6 Conclusions...................................... 82 2.7 Appendix: Attic Reduplication Perfects....................... 84 3 Anatolian 85 3.1 Introduction...................................... 85 3.2 Hittite Orthography, Epenthesis, and Cluster Phonotactics.............. 86 3.3 Hittite and Luwian Reduplication Data........................ 88 3.3.1 CVX Data.................................. 89 3.3.2 VCX Data.................................. 91 3.3.3 TRVX Data................................. 91 3.3.4 STVX Data................................. 92 3.4 Synchronic Analysis of Hittite Copying Patterns................... 92 3.4.1 CVX Bases in Hittite............................ 92 3.4.2 TRVX Bases in Hittite........................... 94 3.4.3 STVX Bases in Hittite........................... 95 3.4.4 VCX Bases in Hittite............................ 98 3.4.5 Hittite Summary............................... 99 3.5 Synchronic Analysis of Luwian Copying Patterns.................. 100 3.5.1 TRVX Bases in Luwian........................... 100 3.5.2 VCX Bases in Luwian, and the Ranking of *PCR............. 100 10
3.6 Reconstructing Proto-Anatolian........................... 101 3.6.1 Reconstructing the Behavior of STVX Bases................ 102 3.6.2 Reconstructing the Behavior of TRVX Bases................ 104 3.6.3 Reconstructing the Absence of VCX Bases................. 106 3.7 Synchronic Analysis of Proto-Anatolian....................... 107 3.8 Constraint Re-ranking and the Demise of *PCR in Anatolian............ 108 3.8.1 The Relative Chronology of Constraint Re-Rankings into Hittite...... 109 3.8.2 Maximally Informative Recursive Constraint Demotion (MIRCD)..... 112 3.8.3 From Proto-Anatolian to Hittite....................... 115 3.8.4 (MI)RCD in Proto-Anatolian......................... 116 3.8.5 From Proto-Anatolian to Luwian....................... 117 3.8.6 MIRCD or a Bias for BR-Faithfulness?................... 119 3.8.7 Additional Support for the Late Demotion of *PCR (in Hittite)....... 120 3.8.8 Hittite šip(p)and- and the Ranking of *PCR before Proto-Anatolian.... 121 3.9 Conclusion...................................... 121 4 Germanic 123 4.1 Introduction...................................... 123 4.2 The Germanic Verbal System............................. 124 4.2.1 Verbal Categories and Inflection in Gothic.................. 124 4.2.2 The Gothic Weak Verbs........................... 125 4.2.3 The Gothic Strong Verbs........................... 126 4.3 Null Morphemes Plus REALIZE MORPHEME.................... 128 4.4 Reconstructing Back to Pre-Proto-Germanic..................... 130 4.4.1 Relevant Sound Changes and Phonological Processes............ 132 4.4.1.1 The Phonemic Merger of */e/ and */i/ in Gothic........ 133 4.4.1.2 Raising Processes in (Pre-)Proto-Germanic............ 133 4.4.1.3 Other Vowel Changes....................... 134 4.4.2 The Strong Verbs in Pre-Proto-Germanic.................. 136 4.4.3 The Vowel System of Pre-Proto-Germanic.................. 137 4.5 Synchronic Derivation of Pre-Proto-Germanic Strong Preterites........... 138 4.5.1 Strong Class I III Preterite Plurals...................... 138 4.5.2 Strong Class IV & V Preterite Plurals.................... 140 4.5.3 Strong Class I V Preterite Singulars..................... 142 4.5.3.1 Strong Class IV V Preterite Singulars.............. 142 4.5.3.2 Strong Class I III Preterite Singulars............... 143 4.5.4 Excursus: REALIZE MORPHEME and Base Priority............. 144 4.5.5 Interim Summary............................... 145 4.5.6 Strong Class VI Preterite Singulars and Plurals............... 145 4.5.6.1 Strong Class VI Preterite Plurals................. 145 4.5.6.2 Strong Class VI Preterite Singulars................ 146 4.5.7 Strong Class VIIa and VIIc Preterite Singulars and Plurals......... 147 4.5.7.1 Strong Class VIIa Preterite Singulars and Plurals......... 147 4.5.7.2 Strong Class VIIc Preterite Singulars and Plurals......... 149 4.5.8 Summary of Analysis............................. 151 4.5.8.1 Stem Formation and the Regular Phonology........... 152 4.5.8.2 Applying the Analysis to Gothic................. 153 11
4.6 Reduplicant Shape and the Stem Formation System................. 155 4.6.1 Synchronic Analysis of Reduplicant Shape (Based on Gothic Patterns)... 156 4.6.2 Evidence of Cluster-Initial Reduplication Patterns in Northwest Germanic. 159 4.7 The Problems with Strong Class VIIb and VIId................... 162 4.7.1 Strong Class VIIb Preterite Singulars and Plurals: A Minor Problem.... 162 4.7.2 Strong Class VIId Preterite Singulars and Plurals: A Bigger Problem.... 163 4.7.2.1 Diachrony of Class VIId: From PIE to Pre-Proto-Germanic... 163 4.7.2.2 Class VIId and the Synchronic System of Pre-Proto-Germanic. 166 4.8 In Defense of Null Morphemes in the Preterite.................... 167 4.8.1 Pre-Requisites for Finding Substantive Underlying Representations.... 168 4.8.2 Arguments for Identifying the Present Stem as the Root.......... 169 4.8.3 The Allomorphy Analysis.......................... 170 4.8.4 Evaluating the Allomorphy Approach.................... 174 4.9 Conclusion...................................... 175 4.10 Appendix I: Strong Verbs in Gothic......................... 176 4.10.1 Strong Class I................................. 176 4.10.2 Strong Class II................................ 177 4.10.3 Strong Class III................................ 177 4.10.4 Strong Class IV................................ 178 4.10.5 Strong Class V................................ 179 4.10.6 Strong Class VI................................ 180 4.10.7 Strong Class VII............................... 181 4.11 Appendix II: Gothic Verbal Inflectional Paradigms................. 182 4.11.1 Weak Verbs.................................. 182 4.11.2 Strong Verbs................................. 184 5 Sanskrit and the C 1 ēc 2 Pattern 185 5.1 Introduction...................................... 185 5.2 Reduplication for Cluster-Initial Roots in Sanskrit.................. 187 5.3 The C 1 ēc 2 Pattern in Sanskrit Zero-Grade Perfects................. 189 5.4 An Allomorphy Analysis of the Sanskrit C 1 ēc 2 Pattern............... 192 5.5 Phonological Analyses of the Sanskrit C 1 ēc 2 Pattern................ 196 5.5.1 A Rule-Ordering Analysis of the C 1 ēc 2 Pattern............... 196 5.5.2 A Stratal OT Analysis of the C 1 ēc 2 Pattern................. 197 5.5.3 A Parallel OT Analysis of the C 1 ēc 2 Pattern................ 199 5.5.3.1 Phonological Interpretation of the Output Form......... 199 5.5.3.2 Candidates and Constraints.................... 200 5.5.3.3 Cluster-Initial Roots within the Parallel Analysis......... 206 5.5.4 The C 1 C 2 ēc 3 Pattern?............................ 208 5.5.5 Local Summary................................ 211 5.6 Vowel Quality and the C 1 ēc 2 Pattern in Sanskrit................... 211 5.7 The Development of the C 1 ēc 2 Pattern in Germanic................. 213 5.7.1 The Evidence from Germanic........................ 213 5.7.2 The C 1 ēc 2 Pattern and Grammar Change in Germanic........... 214 5.8 Conclusion...................................... 216 12
6 The No Poorly-Cued Repetitions Constraint 219 6.1 Introduction...................................... 219 6.2 Empirical Motivation for *PCR............................ 221 6.2.1 Ancient Greek: TRVX C 1 -copying, STVX Non-copying......... 221 6.2.2 Gothic & Proto-Anatolian: TRVX C 1 -copying, STVX Cluster-copying. 222 6.2.3 Sanskrit Cluster-Initial Roots: TRVX C 1 -copying, STVX C 2 -copying.. 223 6.2.4 Sanskrit Zero-Grade Bases: TaR C 1 -copying vs. SaT C 1 ēc 2....... 224 6.2.5 Local Summary................................ 225 6.3 The Cluster-Wise Distributions of Repetition Avoidance Effects.......... 225 6.3.1 Gothic..................................... 226 6.3.2 Greek..................................... 227 6.3.2.1 The Core Facts of Greek...................... 227 6.3.2.2 Voiced Stop + Sonorant Clusters................. 229 6.3.2.3 Fricative + Sonorant Clusters................... 231 6.3.2.4 Nasal-Liquid Clusters....................... 232 6.3.2.5 Apparent Exceptions, and their Ramifications for Pre-Greek... 233 6.3.2.6 Attic Reduplication in Pre-Greek: Avoidance of Repeated Laryngeals................................ 234 6.3.2.7 Local Summary.......................... 234 6.3.3 A Non Indo-European Parallel: Klamath.................. 236 6.3.4 The Distribution of Repetition Avoidance Effects in Sanskrit........ 238 6.3.4.1 Sanskrit Cluster-Initial Roots................... 239 6.3.4.2 Sanskrit Zero-Grade Bases (The CēC Pattern).......... 243 6.3.4.3 Reconciling the Domains..................... 252 6.3.4.4 The Reconciled Distribution of Repetition Avoidance Effects in Later Sanskrit........................... 254 6.3.5 Repetition Licensing and Sonority...................... 256 6.3.5.1 Local Summary.......................... 256 6.3.5.2 Minimum Sonority Distance................... 256 6.3.5.3 Beyond Sonority.......................... 258 6.4 Acoustic/Auditory Cues to Consonantal Contrasts and the Definition of *PCR... 258 6.4.1 Acoustic/Auditory Cues and Consonantal Contrasts............. 259 6.4.2 The Poorly-Cued Repetitions Hypothesis.................. 260 6.4.3 The No Poorly-Cued Repetitions Constraint (*PCR)............ 261 6.4.4 Intensity Rise as the Central Cue for *PCR................. 263 6.4.4.1 Intensity and Sonority....................... 263 6.4.4.2 Intensity Rise and Repetition................... 264 6.4.4.3 Intensity Rise at Release...................... 265 6.4.4.4 Preview of the Typology...................... 266 6.5 Cues and the Language-Specific Definitions of *PCR................ 267 6.5.1 Later Sanskrit................................. 267 6.5.2 Gothic..................................... 269 6.5.3 Rig-Vedic Sanskrit.............................. 270 6.5.4 Klamath.................................... 272 6.5.5 Ancient Greek................................ 273 13
6.5.6 Discussion.................................. 275 6.5.6.1 Assessing the (Micro-)Typology of *PCR Effects........ 275 6.5.6.2 Other Cues and the Repetition Context.............. 275 6.5.6.3 Interim Conclusions........................ 276 6.6 Additional Empirical Evidence for *PCR from Reduplication............ 277 6.6.1 Latin: Infixing Perfect Reduplication in STVX Roots........... 277 6.6.2 Sanskrit: Infixing Desiderative Reduplication in Vowel-Initial Roots.... 278 6.6.2.1 The Position of Infixation and *PCR............... 279 6.6.2.2 Infixation as Cue-Based Faithfulness............... 281 6.6.2.3 Infixation as Templatic Syllable-Alignment............ 285 6.6.2.4 Local Summary.......................... 290 6.6.3 Analysis of the Klamath Distributive..................... 290 6.7 Empirical Evidence for *PCR outside of Reduplication............... 293 6.7.1 Allomorphy in Latin Suffixes in is....................... 294 6.7.2 Aspiration in Sanskrit............................ 296 6.7.3 The *scvc Constraint in English...................... 298 6.7.4 Interim Conclusions............................. 299 6.8 Conclusion...................................... 299 7 Conclusion 301 7.1 Summary of Dissertation............................... 301 7.1.1 Greek..................................... 301 7.1.2 Anatolian................................... 302 7.1.3 Germanic................................... 303 7.1.4 Sanskrit.................................... 303 7.1.5 *PCR..................................... 304 7.2 Discussion of the Methodology............................ 305 7.3 Reconstruction of Proto-Indo-European Reduplication............... 306 References 309 14
Chapter 1 Introduction 1.1 Overview of the Dissertation The reduplicative systems of the ancient Indo-European languages are characterized first and foremost by an unusual, almost unique, alternation in the shape of the reduplicant. The related languages Ancient Greek, Gothic, and Sanskrit share in common the property that root-initial consonant clusters exhibit different reduplicant shapes, depending on their featural composition. Moreover, even though the core featural distinction largely overlaps across the languages, the actual patterns which instantiate that distinction are themselves distinct across the languages. For roots beginning in stop-sonorant clusters (TRVX roots), each of these languages agrees in displaying a prefixal CV reduplicant, where the consonant corresponds to the root-initial stop: TV-TRVX. These three languages likewise agree that roots beginning in sibilant-stop clusters (STVX roots) show some pattern other than the one exhibited by TRVX roots. However, each of the three languages exhibits a distinct alternative pattern: V-STVX in the case of Ancient Greek, STV-STVX in the case of Gothic, TV-STVX in the case of Sanskrit. This dissertation provides an integrated synchronic and diachronic theoretical account of the morphophonological properties of verbal reduplication in the ancient Indo-European languages, with its central focus being to explain this core alternation between TRVX roots and STVX roots an alternation that runs throughout much of the language family, yet consistently differs in such fundamental respects. Set within Base-Reduplicant Correspondence Theory (McCarthy & Prince 1995, 1999, et seq.), a framework for analyzing reduplication in Optimality Theory (Prince & Smolensky 1993/2004), the comprehensive synchronic analyses constructed in service of understanding this distinction and other interrelated distinctions allow us to probe complex theoretical questions regarding the constraints and constraint interactions involved in the determination of reduplicant shape. This dissertation seeks not only to develop in depth, consistent accounts of both the productive and marginal/archaic morphophonological aspects of reduplication in the Indo-European languages, it aims to understand the origins of these patterns from a historical and comparative perspective, and from the perspective of morphophonological learning and grammar change. In so doing, it develops fully articulated synchronic analyses of earlier stages of the languages, and attempts to motivate the conditions for the onset, development, and retention of the changes that result in the systems observed in the attested languages. As such, these analyses constitute a valuable set of case studies on complex systemic change in phonological grammars. Furthermore, having assembled a suite of comprehensive accounts of the reduplicative systems of the individual Indo-European languages, 15
coupled with internal reconstructions based on analysis of their archaic patterns, this dissertation provides new perspective on the comparative reconstruction of the reduplicative system of the common ancestor of the Indo-European languages Proto-Indo-European as a dynamic synchronic system. 1.2 Structure of the Dissertation This dissertation is structured around four in depth case studies in the reduplicative systems of the individual ancient Indo-European languages: Ancient Greek (Chapter 2), the Anatolian languages Hittite and Luwian (Chapter 3), Gothic (Chapter 4), and Sanskrit (Chapter 5). These case studies culminate with the formalization of a new solution to the core TRVX vs. STVX distinction: the NO POORLY-CUED REPETITIONS constraint (*PCR). This approach derives the TRVX vs. STVX distinction which actually only represents the endpoints of a larger, more diverse distributional pattern exhibited within these and other phonological systems (reduplicative and nonreduplicative alike) through stringent contextual licensing conditions on repeated consonants, based on acoustic/auditory cues to consonantal contrast (Chapter 6). Bringing to bear the analytical and historical insights from the individual languages, and the formal mechanism of the *PCR constraint, the dissertation concludes with a new systemic reconstruction of reduplication in Proto- Indo-European (Chapter 7). The case studies largely adhere to a consistent structure and explanatory trajectory. They begin with detailed theoretical analyses of the synchronic reduplicative systems (and the broader morphophonological systems in which they are embedded) as we have them in the attested languages. This analysis reveals some aspect of the system which has become, in one way or another, less than transparent, suggesting that this component of the system may be better understood by considering its context at a prior stage of the language. I identify this prior stage through internal and/or comparative reconstruction, and I systematically examine the properties that underlie the attested synchronic oddity, yielding a new synchronic analysis of the relevant aspects of the prior stage. This raises the question of how the synchronic system of the prior stage, where the relevant component was transparently generated as part of the regular morphophonological system, came to develop into the attested system where this is no longer the case. To address these questions, I consider how principles of (morpho)phonological learning focusing largely on versions of the Recursive Constraint Demotion (RCD; Tesar & Smolensky 1998, 2000; cf. Prince & Tesar 2004, Becker 2009) learning procedure could have driven the change in the phonological and morphological grammar from one stage to the next, deterministically generating the innovative constraint rankings and morphophonological organizations. I now outline the structure and main arguments of the individual chapters. Subsequently, in Section 1.3, I provide an overview of the approach to reduplicant shape that will be employed throughout the dissertation, including a limited factorial typology of the main constraints involved and a preview of the Indo-European reduplication data. 1.2.1 Ancient Greek I begin the body of the dissertation in Chapter 2 with Ancient Greek. In addition to displaying the core Indo-European distinction between TRVX roots and STVX roots in reduplication in the form of a phonologically predictable alternation (driven by *PCR) between C 1 -copying for TRVX roots (TV-TRVX ) and non-copying for STVX roots (V-STVX ) Ancient Greek 16
also shows a synchronically unpredictable distinction in its treatment of vowel-initial roots in the perfect tense, the main reduplicated morphological category of the language. While most vowelinitial roots (VCX roots) form their perfect tense stem via lengthening of the root-initial vowel ( VCX ), a small set of roots instead display a pattern referred to as Attic Reduplication, whereby the root-initial VC sequence is copied and the root-initial vowel is lengthened (VC- VCX ). After constructing an analysis that generates the productive behavior of the consonant-initial roots, I show that this reduplicative grammar is directly compatible with the productive vowellengthening pattern. Attic Reduplication, and also a set of consonant-initial roots which show reduplication in defiance of the normal pattern, can only be generated synchronically through appeal to lexical constraint indexation. The exceptional, lexically restricted behavior of these two types requires additional explanation. Such explanation can be achieved through consideration of their diachronic origins, and how they interact with the rest of the system over the time course of change. I argue that Attic Reduplication can and should be traced back to an alternative reduplication strategy for laryngeal initial roots in Pre-Greek, triggered by laryngeal-related phonotactics. I construct a complete analysis of the Pre-Greek stage, and demonstrate that the properties of the precursor of the Attic Reduplication pattern are shaped by the interaction of the independently motivated reduplicative grammar and another laryngeal-related repairs, laryngeal vocalization. The subsequent loss of the laryngeals via sound change forced learners to reanalyze the pattern, such that Attic Reduplication came to be retained in Ancient Greek not directly by phonotactics, but by lexical constraint indexation. This situation is repeated by the consonant-initial roots that show exceptional copying in reduplication. These forms retain archaic reduplicative properties because of their connection to other morphologically-related reduplicated outputs in such a way that they are brought within the orbit of the lexically indexed constraint that is independently required for Attic Reduplication. The development and synchronic behavior of these types can be motivated and formalized using a procedure for constraint cloning under conditions of ranking inconsistency, based on the proposals of Becker (2009) and Pater (2009). 1.2.2 Anatolian In Chapter 3, I examine and analyze the reduplicative systems of the Anatolian languages Hittite and Luwian, and I undertake a reconstruction of the reduplicative system of their proximate common ancestor Proto-Anatolian. Hittite and Luwian fail to show the core Indo-European distinction between TRVX roots and STVX roots in their synchronic reduplicative systems: Hittite exhibits across-the-board cluster-copying (TRV-TRVX and istv-stvx ), while Luwian lacks the distinction because it has eliminated STVX roots from its lexicon via sound change (it retains C 1 -copying for TRVX roots: TV-TRVX ). Nevertheless, based on the languages regular diachronic correspondences in the treatment of ST clusters, and considerations of parsimony in historical change, Proto-Anatolian is actually to be reconstructed as having had the core TRVX vs. STVX distinction in the form of C 1 -copying for TRVX roots (TV-TRVX ) vs. cluster-copying for STVX roots (STV-STVX ). The synchronic activity of *PCR is essential in generating the TRVX vs. STVX distinction in Proto-Anatolian. However, this constraint plays no role in the synchronic grammar of its daughter languages Hittite or Luwian; in fact, the two languages bear an innovative pattern of reduplication for vowel-initial roots (VR-VRT ) that directly violates it. Therefore, *PCR must transition from a high-ranked, active constraint which can induce alternative reduplication strategies in Proto-Anatolian to a low-ranked, inactive constraint which can be violated in reduplication in the daughter languages. I will argue that independent phonological changes in the internal development of Hittite and Luwian had the incidental effect of eliminating the distinction between TRVX 17
roots and STVX roots in reduplication, and that this allowed learners to converge on a *PCR-free analysis, paving the way for the subsequent emergence of the *PCR-violating VR-VRT pattern. This represents an interesting case with respect to phonological learning, as it seems to be a diachronic counter-example to the Subset Principle (cf. Prince & Tesar 2004). That is, there is a point in the development of these languages where speakers evidently did not learn the most restrictive grammar, allowing for the later emergence of a more marked pattern. Given the particular constraint types that are involved, it might be possible to avoid the problem by using Biased Constraint Demotion (BCD; Prince & Tesar 2004) as the learning procedure. Alternatively, the problem may be solved by introducing a minor reformulation of RCD/BCD that favors the high ranking of maximally informative winner-preferring constraints. This approach yields a satisfactory step-wise account of the relevant diachronic developments, and it would support a view that, under the right conditions, learners will indeed be predicted to learn a non-subset grammar. 1.2.3 Gothic In Chapter 4, I explore the analysis of the reduplicative system of Gothic. Unlike the other languages examined in detail in this dissertation, the reduplicated verbal forms of Gothic represent but one piece of a larger system of complex verbal morphology and morphophonology the so-called strong verb system. The preterite stems of strong verbs are formed in a variety of ways, mostly involving vocalic alternations relative the base, but also including reduplication. What is especially noteworthy about this system is that the choice of stem-formation process is predictable based on the phonological properties of the verbal root; for example, reduplication is only found in roots having a long root vowel or a root vowel /a/ followed by two consonants. Therefore, the distribution of preterite stem formation patterns seems very clearly non-arbitrary from the synchronic point of view, and is thus in need of analysis. I develop an analysis whereby the strong preterites select for a null underlying representation of the morpheme PRETERITE, and differentiation of stems is induced by a family of constraints that require overt exponence of morphosyntactic features, REALIZE MORPHEME (RM; Kurisu 2001), and thus phonological contrast between stems which are morphologically related in a particular way. The nature of the changes undergone to satisfy RM, of which reduplication is only one of many (and indeed a sort of last resort synchronically), falls out from the interaction between the phonological properties of individual roots and the ranking of markedness and faithfulness constraints. Furthermore, I show that the phonological grammar required to generate these patterns of stem formation are fully consistent with the phonological grammar required to account for the patterns of reduplicant shape, which are themselves subject the core TRVX vs. STVX distinction induced by *PCR. 1.2.4 Sanskrit Chapter 5 examines certain aspects of the complex reduplicative system of Sanskrit. The perfect tense in Sanskrit demonstrates two distinct means of instantiating the core Indo-European TRVX vs. STVX distinction. For cluster-initial roots, this distinction is reflected in the difference between C 1 -copying for TRVX roots (TV-TRVX ) and C 2 -copying for STVX roots (TV-STVX ). On the other hand, in the inflectional categories of the perfect which normally call for deletion of the root vowel ( zero-grade ablaut ), collectively referred to as the perfect weak stem, the TR vs. ST distinction plays out in a different way for roots of the shape C 1 ac 2, which would be reduced to C 1 C 2 clusters by root-vowel deletion. TaR roots show the expected root-vowel deletion and exhibit C 1 -copying: i.e., TaR perfect weak stem TV-TR. SaT roots, on the other hand, display neither 18
root-vowel deletion nor reduplication of any kind (at least overtly), instead showing a root allomorph in C 1 ēc 2 : i.e., SaT perfect weak stem SēT. In this chapter, I will develop an analysis of Sanskrit that simultaneously generates each of the distinct outcomes of the perfect, namely: (i) the C 1 -copying pattern for TRVX roots and TaR perfect weak stems, (ii) the C 2 -copying pattern for STVX roots, and (iii) the C 1 ēc 2 pattern for SaT perfect weak stems. I will first present an analysis based on allomorph selection, where the underlying representation leading to the C 1 ēc 2 form is selected in the phonological component just in case a licit reduplicative structure that satisfies *PCR and the Input-Reduplicant faithfulness constraint LINEARITY-IR (McCarthy & Prince 1995, 1999) cannot be obtained. I will then explore an analysis which derives the C 1 ēc 2 pattern directly in the phonology, treating it as a *PCR-driven phonological repair of an overtly reduplicative structure; specifically, //C 1 e-c 1 C 2 -// [C 1 e:c 2 -], where C 1 represents the consonant of the reduplicant, the vowel represents the vowel of the reduplicant, and the length of the vowel derives from compensatory lengthening after the deletion of the root-initial consonant. While the phonological solution encounters difficulties in accounting for the quality of the vowel synchronically in Sanskrit (suggesting that the allomorphy analysis is most appropriate for Sanskrit synchronically), the analysis extends seamlessly to a related C 1 ēc 2 pattern found in the Germanic languages (namely, the Class V strong preterite plurals). 1 I adduce evidence for similar patterns in Old Irish and in Hittite, as well. Regardless of the synchronic analysis of the pattern in attested Sanskrit, this correspondence argues in favor of a phonological origin of the C 1 ēc 2 type, even if they do not originate as a unitary formation in Proto-Indo-European. 1.2.5 The NO POORLY-CUED REPETITIONS Constraint (*PCR) The analyses of the core Indo-European TRVX vs. STVX distinction developed in Chapters 2 5 are centered around the operation of a simplified version of the NO POORLY-CUED REPETITIONS constraint (*PCR), which prohibits sequences of repeated consonants (C α VC α ) in immediate preobstruent position (/_C [-sonorant] ). Chapter 6 expands the scope of this examination so as to include the full cluster-wise distribution of default vs. alternative pattern reduplication in the languages discussed. In each of these reduplicative systems, the TRVX vs. STVX distinction instantiates a broader distributional pattern regarding the treatment of different types of root/base-initial clusters; however, the languages diverge with respect to which clusters pattern with TRVX in allowing the default C 1 -copying pattern, and which clusters pattern with STVX in resorting to the alternative pattern. For example, for stop-sibilant (TSVX ) roots, Ancient Greek exhibits the alternative pattern (grouping together with STVX ), while Sanskrit exhibits the default pattern (grouping together with TRVX ). I argue that the full cluster-wise distributions of repetition avoidance effects in these systems are not to be explained in terms of traditional phonological features or abstract phonological properties like sonority, but rather in terms acoustic/auditory cues (cf. Steriade 1994, 1997, 1999, Flemming 1995/2002). Motivated by the empirical evidence, I propose the POORLY-CUED REPE- TITIONS HYPOTHESIS previewed in (1) which states that repetition imposes distinct burdens on the perceptual system with regards to the licensing of contrasts (specifically C Ø contrasts). Based upon this hypothesis, I formulate the final, precise version of the NO POORLY-CUED REPE- TITIONS constraint (*PCR) previewed in (2) which penalizes repetitions that would leave the repeated consonant without sufficient cues to its contrast with Ø. Specifically, I will show that intensity rise, alongside transitions and (perhaps) stop release burst, are central to licensing these 1 Note, however, that the correspondence in vowel quality between the two patterns is illusory: Germanic ē corresponds with Sanskrit ā. 19
contrasts under repetition. The different languages in effect select from among these cues which ones will be sufficient, either on their own or in combination, to license a consonant repetition. (1) THE POORLY-CUED REPETITIONS HYPOTHESIS There is some property of the perceptual system which degrades listeners ability to apprehend the presence of a consonant (i.e. the contrast between that consonant and its absence) when that consonant is adjacent to an identical consonant. i. This property diminishes the effectiveness of some or all acoustic/auditory cues to C Ø contrasts, such that some cues which are normally sufficient to license those C Ø contrasts (in otherwise equivalent positions) are no longer sufficient to license those contrasts under repetition. ii. This property diminishes the effectiveness of different cues to different extents: the effectiveness of cues to acoustic events which are more difficult to anchor at a particular point in the speech stream and/or tend to extend across multiple segments are diminished to a greater degree than cues to acoustic events which are more reliably located at their correct position in the speech stream. (2) The NO POORLY-CUED REPETITIONS constraint (*PCR) Languages may set stricter conditions (in terms of cues) for the licensing of C Ø contrasts (i.e. the presence of C) when that C would be the second member of a transvocalic consonant repetition (i.e. C 2 α in a C 1 αvc 2 α sequence) than in other contexts. Assign a violation mark * for each C 2 α (i.e. each C Ø contrast where C is a C 2 α) which is not cued to the level required by the language-specific repetition licensing conditions. The inclusion of this constraint in the grammar properly derives the patterns of repetition avoidance instantiated by the languages examined in this dissertation. In Chapter 6, I adduce additional empirical and analytical evidence in favor of the use of the *PCR constraint. I provide analyses of several additional reduplicative effects that require the use of *PCR, namely, infixal reduplication in Latin STVX roots, infixal reduplication in Sanskrit vowel-initial desideratives, and a pattern equivalent to that of Gothic found in the non Indo-European language Klamath (Barker 1964). I also discuss a few marginal patterns outside of reduplication which may be amenable to a *PCR-based analysis. 1.2.6 Reconstructing Proto-Indo-European Reduplication Chapter 7 concludes the dissertation. Besides reviewing the main arguments advanced in the earlier chapters, this chapter lays out an updated reconstruction of the reduplicative system of Proto-Indo- European (PIE), focusing on the question of the core TRVX vs. STVX distinction. The more traditional view reconstructs for PIE the distribution found in Gothic (and now Proto-Anatolian): i.e., C 1 -copying for TRVX (TV-TRVX ) but cluster-copying for STVX (STV-STVX ). This allows for the productive STVX treatments of Ancient Greek (V-STVX ), Sanskrit (TV-STVX ), and Latin (S-TV-TVX ) to be seen as different kinds of reductions from the original type. I argue for an alternative reconstruction for PIE, one which is present in the literature (e.g., Byrd 2010:100 105), though not the most commonly accepted view: in PIE, both TRVX and STVX roots, and indeed all types of root-initial clusters, exhibited C 1 -copying; i.e., TV-TRVX and SV-STVX. The strongest evidence in favor of this alternative reconstruction comes from archaisms in Greek and Latin (Brugmann & Delbrück 1897 1916:40 41, Byrd 2010:103 104), and their agreement with Iranian (Byrd 2010:103), which all points to SV-STVX. 20